The Generic Ballot
A single question can give us a sense of who is winning the race for the House
Polling is expensive and to do it right corners cannot be cut. For any poll to be both scientific and useful, a representative sample of at least 400 respondents is necessary. It is much better to have a sample of 600 or more. The cost for one such poll could be as little as $5,000, but it is usually more than that – especially if you want a bigger sample size (which, as you know by now, reduces the margin of error).
There are 435 members of the House of Representatives and they are all up for election every two years. That means, of course, that there are 435 elections for the House taking place right now. At the least, that’s a price tag of over $2 million.
The “generic ballot” is one way to get a sense of which party is likely to win the House without having to survey all or even most of the 435 districts. The question is asked by some, not all, pollsters and it is simple. The question is typically posed like this: “If the election were held today, would you vote for the Democrat or Republican candidate* for Congress?”**
The generic ballot cannot tell us who is likely winning in any particular district, but historically it does a fairly good job of predicting who will win the House and by what magnitude. According to Pew, “[t]he final forecast of the generic House vote and the actual vote totals have paralleled each other very closely for nearly a half-century in U.S. elections. The average prediction error in off-year elections since 1954 has been 1.1%.” Although Pew argued in 1996 that the generic ballot was less reliable during presidential years, since then it has become as reliable as it is in the mid-terms. Perhaps this is due to the hyperpartisanship introduced by Gingrich in the mid-90s which has led to the reduction of ticket-splitting. In fact, as the table below illustrates, Nate Silver found that since 1996 the generic ballot has done a better job predicting the final House margin in presidential years (two point error) than in mid-terms (three point error).
Let’s take a look at the current state of the generic ballot. There has been an uptick in surveys reporting this metric recently. In the past week, a simple average of the 12 polls reporting findings on this question shows the Democrats leading by +7.4 points. Simply using the average two point error margin, this means that the Democrats should win the House (technically, they should win the most votes – which is not the same thing) by between five and ten points. The generic ballot is also considered an indicator of the national mood, particularly in this age of few ticket-splitters. So the margin for the Democrats here is also a good sign – and another indicator – that Biden is winning the election right now.
A couple of House races to watch
Here are a couple of Congressional races to watch that might give us an idea of how big the Democrats’ night could be on November 3rd. In Texas-10, which covers parts of Austin and some suburbs of Houston, the incumbent Republican Rep. Michael McCaul is leading his Democratic opponent, Mike Siegel, by two points. This is a district that as a Partisan Voter Index (PVI) rating of Republican +9 (R+9). This is a metric used by Cook Political Report that shows how much more Republican or Democratic a district votes in presidential elections compared to how the national electorate typically votes. This means that a Republican presidential candidate typically gets a vote share that is nine points higher in TX-10 than his or her party gets nationwide. That McCaul is currently up by two points tends to validate the generic ballot average right now. If the Democrats win the House vote nationally by seven points and the voters of TX-10 give the GOP candidate a nine point advantage, McCaul will win by two points. THat’s not precisely what the PVI tells us, but it’s a useful enough assumption considering how many people vote a straight party-line ticket today. But, what it also means is that this race is competitive and can easily (this year) be picked up by the Democrats with a little help.
The other district is Pennsylvania-01. This is an interesting race to watch because it has a PVI of R+1 and, not surprisingly, nearly equal numbers of Democratic and Republican registered voters. This race may tell us something about how the national election will break in close districts. The Republican incumbent, Brian Fitzpatrick, is leading his Democratic opponent, Christina Finello, by one point. To be sure, there is very little polling in PA-01 and TX-10 (each had a poll this week, but just one each), so we cannot be sure that either poll is accurate because there are not other polls to compare them to. CPR rates both districts as Lean Republican.
* Good pollsters will alternate the party names to avoid any bias that might attach to hearing the first (or last) name used.
** Americans often use “Congress” to mean “House of Representative” even though Congress is the official name of the national legislature consisting of both the Senate and the House of Representatives. In fact, representatives are routinely referred to as Congressmen and Congresswomen. Senators are never referred to that way.
Why is there little polling in Texas? I had no idea they cost thousands, wow!